changing scales in output for vector fields causes simulations to crash

[jsoishi]

If an analysis task including a vector field scaled to greater than 1,

hires=True
if hires:
    logger.info("saving high resolution data")
    hires = solver.evaluator.add_file_handler('hires', sim_dt=0.25, max_writes=50)
    hires.add_task(u, name='u_scales', scales=4)

the simulation will crash immediately after the first write. This is true even if it is the only task in the file handler. Scalar fields do not have this problem:

hires.add_task(b, name='b_scales', scales=4)

works fine.

For a minimum failing example, add the lines above to the rayleigh_benard.py test problem. This problem occurs in both parallel and serial.

I'm fairly certain this is a regression since mid-November 2023, but I have not had a chance to go back and bisect the code.